BullGuru

• 

  Starting screen where you set up atbat.

• 

  An interactive strike zone with dropdowns to submit pitch results, paired with pitch recommendations displayed in the left sidebar.

Inspiration

Growing up with immigrant parents, spending large amounts of money on sports was never an option. I watched peers with access to private coaches and elite facilities develop much faster—not just physically, but mentally. A good coach does not only refine mechanics; they teach pitch sequencing, game awareness, and how to think several steps ahead. When I was pitching at a young age, I threw to random locations and chose pitches based on feel rather than intentionally setting up hitters for a strikeout. I built BullGuru to fill that gap by acting as a virtual pitching coach and strengthening a young pitcher’s mental approach to the game.

What It Does

BullGuru recommends the optimal next pitch during an at-bat based on the batter’s handedness, the pitcher’s handedness, the pitcher’s pitch arsenal, the current count, and the outcomes of previous pitches in the at-bat.

How I Built It

I extracted and processed MLB strikeout data using Python to identify effective pitch sequencing patterns. I trained a lightweight model and exported the results into a format suitable for real-time inference. The application is built with a modern web stack using Node.js and a Next.js frontend, with serverless API routes that deliver fast, low-latency pitch recommendations.

Challenges I Ran Into

One of the biggest challenges was transforming noisy, real-world MLB pitch data into clean, structured sequences that accurately reflected decision-making rather than raw outcomes. Balancing model complexity with real-time performance was also critical to keep the application fast and accessible.

Accomplishments I’m Proud Of

I built an end-to-end system that combines professional baseball data, machine learning, and a user-friendly web interface. BullGuru delivers meaningful, coach-like guidance without requiring expensive hardware or recurring subscriptions.

What I Learned

This project taught me how to design data pipelines, train and deploy lightweight machine learning models, and integrate them into a full-stack web application. I also learned how to build products that directly address accessibility problems.

What’s Next for BullGuru

My goal is to turn BullGuru into a one-time, low-cost mobile app (around \$3) to make it accessible to young players everywhere. I also plan to expand the platform with computer vision features to identify professional pitcher comparisons, analyze pitching mechanics, and generate AI-driven warm-up and stretching routines based on a player’s flexibility.

Built With

• css
• dropdowns
• for
• html5
• interactive-svg-for-strike-zone
• javascript-frontend-framework:-react
• json
• json-model-export-cloud-/-deployment:-vercel-(for-hosting-and-serverless-functions)-apis:-none-external-(ml-recommendations-generated-from-processed-mlb-data)-database-/-storage:-json-files-for-model-data-and-pitch-sequences-other-technologies:-html/css
• languages:-python
• next.js-backend-/-server:-node.js
• pitch
• react.js
• serverless-api-routes-machine-learning-/-data:-python-ml-scripts-for-pitch-sequence-modeling
• svg
• vercel